Announcement

Collapse
No announcement yet.

The AI cheats on Noble, correct?

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #31
    Originally posted by Arrian
    I actually think it will be pretty funny if and when someone finally produces a smart Civ AI that beats humans w/o cheating.

    There will be much wailing and gnashing of teeth.

    -Arrian
    Yes, it's amazing how much people will find to complain about as soon as there is nothing to blame their loosing on.

    Tom P.

    Comment


    • #32
      Using the chess analogy:

      A chess board has a finite number of moves, finite number of pieces, and centuries of known and improved-upon strategies that can be programmed into a computer. Ironically, the best way to beat a chess computer is to do something that it doesn't understand in the number of move computations it makes before moving.

      32 pieces on 64 squares is easier to do than 400 different pieces on 100x100 squares.

      Also CIV has a mechanism that chess doesn't have as much (pawn to the back row yes, but it doesn't create units, just promotes them), the option to create more units. because there is a finite amount of pieces in chess, strategy itself is different than in CIV where you could just max out production and make axemen for example.

      To create an AI like the one being described would not even be possible for Civ I let alone CIV right now. Someday I'm sure I'll eat my words, but the AI that CIV has, is probably the best AI that game could have gotten without isolating too many customers with slow "waiting for AI Civilization" things etc.
      First Master, Banan-Abbot of the Nana-stary, and Arch-Nan of the Order of the Sacred Banana.
      Marathon, the reason my friends and I have been playing the same hotseat game since 2006...

      Comment


      • #33
        Originally posted by NFIH
        I don't know how to interpret all the numbers there since I have no baselines to compare them to and anyway wouldn't know which numbers are significant and which are not.
        Here's a rather useful chart that has been passed around lately. Take a look if you haven't already. It helps get a better understanding of the handicap stuff: http://www.apolyton.net/forums/showt...hreadid=148571
        One of these days I'll make 501 posts, and you won't have to look at my silly little diplomat anymore.
        "Oh my God, what a fabulous room. Are all these your guitars?"

        Comment


        • #34
          I understand that Civ 4 is not comparable to chess and wrote so in my first post. And I am aware that there is no alternative to hefty material advantages, general map knowledge etc. right now. Nonetheless it feels not right in a TBS and I understand the feelings of those who protest against it, well knowing that there is no other way in sight.

          I would probably prefer a much limited Civ version (less tiles, less units, simpler rules) with an AI that does not cheat. I might even try and start such a project. Not for release, as I suck at graphical programming. It would have a rather Nethack-esque interface. My only premise would be, that it does not cheat.

          Comment


          • #35
            ....and the human would always win and get bored quite quickly.
            It's almost as if all his overconfident, absolutist assertions were spoonfed to him by a trusted website or subreddit. Sheeple
            RIP Tony Bogey & Baron O

            Comment


            • #36
              Originally posted by rah
              ....and the human would always win and get bored quite quickly.
              What part of "less tiles, less units, simpler rules" did you not understand?

              Chess programs can beat grandmasters today. An overwhelming majority of players aren't grandmasters. So chess programs pretty much could pwn 99.9% of their opponents already today.

              Most Civ players are grandmasters either. Except maybe Aeson. So the situation is similar.

              Why should it be impossible to find a ruleset on a, say 64x64 or even a 32x32 "board" with 8-10 different unit types? This would get that game closer to chess dimensions, with which modern computers very well can cope.

              Comment


              • #37
                Fritz 8 kicks my butt almost every time, and I'm no slouch when it comes to chess.
                One of these days I'll make 501 posts, and you won't have to look at my silly little diplomat anymore.
                "Oh my God, what a fabulous room. Are all these your guitars?"

                Comment


                • #38
                  Originally posted by NFIH
                  I don't know how to interpret all the numbers there since I have no baselines to compare them to and anyway wouldn't know which numbers are significant and which are not.
                  You can compare the numbers for the human player between various levels and to the "AI" numbers of the same level. That will give you an idea of how to modify them.

                  Originally posted by NFIH
                  At that particular point in the game, I couldn't have four or five cities AND a large army. It would have to be one or the other. But my point is the AI appeared to have both as many cities, a large army and was also keeping up with me in research. Sounds fishy to me.
                  Not necessarily. It's not that difficult on Noble to have a bunch of cities, a large army, and good research. You just need a bit more planning and knowledge. Right now I play on Prince/Monarch (I am not Blake, let alone Aeson, but then again I don't want to think too much ), and it's still not difficult to be ahead of the computer players all the way through. Sure, my army may not be the biggest, but it is big enough to deter all would be attackers*.



                  * Except for people such as Monty or Isa.
                  (\__/) 07/07/1937 - Never forget
                  (='.'=) "Claims demand evidence; extraordinary claims demand extraordinary evidence." -- Carl Sagan
                  (")_(") "Starting the fire from within."

                  Comment


                  • #39
                    Originally posted by padillah
                    Even parallel processing isn't going to help unless they tie the processes together in some meaningful way.
                    Well it would be for strategic games such as Civ. For example, suppose we build a custom computer so that each tile of the map could be analysed in parallel, even for the largest map. Then it would be much easier to write a tougher AI that operates in meaningful time*.

                    Don't forget, Deep Blue is a massively parallel supercomputer.




                    * Players are willing to wait for chess programs to take 15 - 20 minutes to make a move, but not for Civ.
                    (\__/) 07/07/1937 - Never forget
                    (='.'=) "Claims demand evidence; extraordinary claims demand extraordinary evidence." -- Carl Sagan
                    (")_(") "Starting the fire from within."

                    Comment


                    • #40
                      Originally posted by Urban Ranger


                      Well it would be for strategic games such as Civ. For example, suppose we build a custom computer so that each tile of the map could be analysed in parallel, even for the largest map. Then it would be much easier to write a tougher AI that operates in meaningful time*.

                      Don't forget, Deep Blue is a massively parallel supercomputer.




                      * Players are willing to wait for chess programs to take 15 - 20 minutes to make a move, but not for Civ.
                      I stand corrected. Massive parrallel processing would help a great deal.

                      This could very well be done now, except for the meaningful response time. The same thing happens in Pixel Shading, each pixel is given a short kind of formula to process and as the POV changes the input to that formula change thus the screen changes - the same could be done for the tiles. Maybe even an pre-evaluation formula to determine if you are empty ocean or unremarkable in some way (grassland with no units on it) then it could skip processing unless it was specifically asked to process by a neighboring tile (for Plan of Attack type stuff).

                      Or, maybe better, each leader has an array of "tiles of interest" and before they make a decision they pre-process the tiles of interest. Now when they make a decision they have a weighted response that can accept feedback (if it's done right).

                      As I think about it a bigger problem would be feedback. We can recall and associate mistakes with consequences, how would the AI? Being such a long, drawn-out game there may be no correlary between their loss and the moves they made. They could have had a bad start or got iron too late or one of a thousand things.

                      So how do we correct the "bad" behavior without understanding how it was bad and why? Or even what was bad and what wasn't.

                      I think the "tiles of interest" idea is still a good one and you're right, it would help. But I don't see a way around a pre-programmed AI. Not for a game as deep and convoluted as this.

                      Tom P.

                      Comment


                      • #41
                        Originally posted by NFIH


                        OK, but how many cities did you have relative to the AI? You had as many or more cities than the AI and a large army and a tech lead?
                        Get a monopoly tech, trade it off to several opponents at once and you'll easily gain a lot of techs if you feel like it.. and the one thing i learnt in my first cIV game, which the AI usually does a lot of, build cottages! lots!
                        Proud member of the PNY Brigade
                        Also a proud member of the The Glory Of War team on PtW-DG

                        A.D 300, after 5h of playing DonHomer said: "looks like civ2 could be a good way to kill time if i can get the hang of it :P"

                        Comment


                        • #42
                          The AI gets advantages with upgrades, unit maintenance, and combat vs. barbarians and animals on Noble, that's it. Well, I think they MIGHT have some kind of advantage on inflation, but I'm not sure, and I'm not even sure what role inflation plays in the game.

                          Comment


                          • #43
                            Here is a 1-page MS Word chart of all the relevant variables of ALL the difficulty levels, Noble level hilighted:
                            Attached Files

                            Comment


                            • #44
                              OK, what kind of variable name is iStartingLocPercent?

                              The percentage chance the player will start at a location?

                              My other favorite is in CIV4BonusInfos.xml, iConstAppearance. Well, if it's a CONST why can I change it?! What the heck does that even mean?

                              ...or bArea?

                              They have the weirdest variable names I've ever seen.

                              Tom P.

                              Comment


                              • #45
                                padillah, I didn't say I understood them all!

                                iStartingLocPercent is obviously something to do with relative starting locations.

                                I haven't looked at that other file yet. Obviously, that CONST is not referring to 'constant' (unless in reference to something else?). Undocumented variables are always the bane of understanding, but they have done fairly well with the names, in general. Just imagine, it could have been named 'cstap'.

                                Check out XML files - Civ4Wiki . While it says both of the vars you mentioned are 'Unknown', you can be confident that if the var's are always 100 or 0 that they are somewhat superfluous. Hmm, maybe that CONST was referring to 'constant' after all.

                                Comment

                                Working...
                                X